People ask this as if there's a universal number. There isn't. There is a useful default, and then a short list of cases where you should slow down or speed up.

The useful default

About every 30 seconds for anything customers pay for or sign into. Checkout, login, the API your mobile app hits, the webhook receiver that completes orders. A five-minute interval means an outage can run for five minutes before the monitor even notices, plus whatever confirmation logic you add on top. For a product that lives online, that window is usually too long.

About every five minutes for marketing sites, docs, staging, and internal tools nobody is refreshing at 2am. You still want to know when they die. You do not need to know in thirty seconds.

Anything slower than five minutes for a live product is mostly theater. You will find out from a customer first.

What the interval actually buys you

Check frequency is time-to-detection, not "quality of monitoring." A thirty-second check with a single flaky probe will page you more often and less accurately than a thirty-second check that waits for regions to agree. Speed without evidence is just a louder pager.

So when someone says "we need ten-second checks," ask what they will do with the extra twenty seconds, and whether their confirmation logic can keep up. Many teams that chase sub-minute intervals are really trying to fix false silence from a five-minute free plan — a real problem — by buying speed they have not earned with trust.

A simple way to choose

Ask one question per URL: If this broke for ten minutes and I only found out at the end, would I be angry?

• Yes → thirty seconds (ten on Pro if the budget is that tight).
• Mildly annoyed → five minutes.
• I would shrug → maybe you do not need a monitor on it yet.

Then apply the same rule to confirmation. If you page on the first failed probe from one place, a fast interval will teach your team to mute alerts. If you wait for several regions to agree, you can keep the fast interval without turning the night into noise.

What I run in practice

On Tallwatch, free monitors check every 30 seconds from three regions with consensus before a page. That is intentional: the free plan is meant to be something you can leave on a real side project, not a five-minute hobby checker you outgrow the first week a customer pays you. Paid plans go to ten seconds when the product genuinely needs it — not because shorter always looks better on a pricing table.

If you are choosing an interval today, start at thirty seconds for the three URLs that would hurt if they died, five minutes for everything else, and spend your energy on when a failure becomes a page, not on shaving seconds off a check that still trusts a single opinion.